Airbnb object recognition

Creates a collection of JSON files with recognized objects in photos of Airbnb listings.

Obtaining the photos

Use our scrape-airbnb tool. This will also pull in all other data about the listings. To run the object recognition on the photos, place them in a folder called photos with the following subfolders:

• host-photos containing the picture of the host of each listing, named [listing-id].jpg
• listing-photos containing subfolders for each listing, which in turn contain all photos of that listing, starting at 0.jpg

Running the script

We use Darknet's Yolo for object recognition. To configure this, do the following:

# Download the darknet repo
git submodule update --init --recursive

# Compile the executable
cd darknet
make

# Download pre-trained weight file
wget https://pjreddie.com/media/files/yolov3.weights

Once Darknet is configured, you can run the script with:

python recognition.py

About this project

airbnb-object-recognition and scrape-airbnb were written as part of a workshop at Non-Linear Narrative at Royal Academy of Art The Hague. They are tools used to create No Home Like Place.